Unable to start contiv-vpp due to SIGTERM

[deepakgunjal]

I am trying to run the contiv-vpp on a single node kubernetes cluster with version v1.14.1 on an ubuntu 16.04 server. This server has a Mellanox card "37:00.0 Ethernet controller: Mellanox Technologies MT27800 Family [ConnectX-5]".

The file /etc/vpp/contiv-vswitch.conf: unix { nodaemon cli-listen /run/vpp/cli.sock cli-no-pager coredump-size unlimited full-coredump poll-sleep-usec 100 } nat { endpoint-dependent translation hash buckets 1048576 translation hash memory 268435456 user hash buckets 1024 max translations per user 10000 } acl-plugin { use tuple merge 0 } dpdk { dev 0000:37:00.0 uio-driver vfio-pci } api-trace { on nitems 5000 } socksvr { default } statseg { default }

cat /proc/meminfo | grep Huge AnonHugePages: 0 kB HugePages_Total: 1024 HugePages_Free: 1016 HugePages_Rsvd: 63 HugePages_Surp: 0 Hugepagesize: 2048 kB

The contiv-vpp pod is not starting up and I am getting the following error:

time="2019-05-17 08:18:19.74759" level=debug msg="/usr/bin/vpp[28474]: load_one_vat_plugin:67: Loaded plugin: vmxnet3_test_plugin.so\n" loc="contiv-init/vpplogger.go(23)" logger=vpp time="2019-05-17 08:18:19.75611" level=debug msg="/usr/bin/vpp[28474]: load_one_vat_plugin:67: Loaded plugin: acl_test_plugin.so\n" loc="contiv-init/vpplogger.go(23)" logger=vpp time="2019-05-17 08:18:19.76391" level=debug msg="/usr/bin/vpp[28474]: load_one_vat_plugin:67: Loaded plugin: mactime_test_plugin.so\n" loc="contiv-init/vpplogger.go(23)" logger=vpp time="2019-05-17 08:18:19.77169" level=debug msg="/usr/bin/vpp[28474]: load_one_vat_plugin:67: Loaded plugin: lacp_test_plugin.so\n" loc="contiv-init/vpplogger.go(23)" logger=vpp time="2019-05-17 08:18:19.87490" level=debug msg="/usr/bin/vpp[28474]: dpdk: EAL init args: -c 2 -n 4 --in-memory --file-prefix vpp -w 0000:37:00.0 --master-lcore 1 \n" loc="contiv-init/vpplogger.go(23)" logger=vpp time="2019-05-17 08:18:27.41260" level=error msg="VPP terminated, stopping contiv-agent" loc="contiv-init/main.go(440)" logger=contiv-init time="2019-05-17 08:18:27.41296" level=info msg="Signal terminated received during agent start, stopping" loc="agent/agent.go(152)" logger=agent time="2019-05-17 08:18:27.42598" level=debug msg=exiting loc="contiv-init/main.go(453)" logger=contiv-init

Please help to understand what could be the cause of this error.

[rastislavs]

Hi, please read hugepages section in https://github.com/contiv/vpp/blob/master/docs/setup/MANUAL_INSTALL.md#hugepages-kubernetes-110-and-above

In k8s 1.14, disabling HugePages fature gate is not supported anymore, so you may need to define memory limit for the vSwitch container:

    resources:
      limits:
        hugepages-2Mi: 1024Mi
        memory: 1024Mi

[deepakgunjal]

Thanks for suggestion. I did that but also found the contiv-stn running which I stopped. Now the VPP is running.

Though there is one observation that since it is a single node cluster and physical NIC is not connected with any cable, the vppctl "show int" command does not show the main VPP interface as given in the installation guide. The output as of now is,

vpp# show int Name Idx State MTU (L3/IP4/IP6/MPLS) Counter Count local0 0 down 0/0/0/0 loop0 1 up 9000/0/0/0 loop1 2 up 9000/0/0/0 loop2 4 up 9000/0/0/0 tap0 3 up 1450/0/0/0 rx packets 157 rx bytes 55900 tx packets 160 tx bytes 16214 drops 9 ip4 148 ip6 8 tap1 5 up 1450/0/0/0 rx packets 138 rx bytes 12365 tx packets 73 tx bytes 27537 drops 62 ip4 130 ip6 8

If I join a new minion in this cluster would any pod on that server be able to connect with the pod on master provided I get physical cable also connected on master node VPP NIC?

[rastislavs]

Hi,

so is your DPDK setup correct according to this manual? https://github.com/contiv/vpp/blob/master/docs/setup/VPP_CONFIG.md

Also, make sure that required kernel modules are loaded: https://github.com/contiv/vpp/blob/master/docs/setup/MANUAL_INSTALL.md#setting-up-dpdk

In case that the node interconnect interface still does not appear on VPP, please provide the logs from the contiv-vswitch pod.